Tool Types

What can AI do?

A tool type is the job the AI is trying to do. Chat is not the same as video. Voice is not the same as music. Connectors are a bigger privacy decision than a simple prompt.

A dark desk with AI stations for chat, image, video, voice, music, connectors and agents

The shelf

Choose by job, not by hype.

If you are new to AI, start with one small job. Ask it to explain, draft, sort, compare or make a harmless image. Do not connect private accounts or upload sensitive files until you understand the terms.

Super apps

One app that does many jobs

ChatGPT, Gemini, Claude and Copilot are becoming AI control rooms: text, image, voice, files, search, tools and connectors in one place.

Good forTrying AI without learning ten apps.

Watch forData sharing, account connections, paid tiers and hidden complexity.

LLMs

Text brains that predict and reason

Large language models write, explain, summarise, translate, plan and help with code. They can sound confident while wrong.

Good forPlain-English explanations, drafts, study help and thinking support.

Watch forHallucinations, missing context and advice that sounds more certain than it is.

Image

Pictures from words or references

Tools like Midjourney, Imagen, DALL-E, Firefly and Stable Diffusion turn prompts into images or edit existing pictures.

Good forStoryboards, posters, mockups, mood boards and visual learning.

Watch forCopyright, likeness, stereotypes, public galleries and fake evidence.

Video

Moving images and clips

Tools like Sora, Veo, Runway, Pika and Kling can make short moving scenes, extend shots or animate images.

Good forConcept clips, film ideas, explainers and social media drafts.

Watch forDeepfakes, copyright, cost per generation and polished-looking false scenes.

Voice

Speech, cloning and dubbing

Voice AI can read scripts, translate speech, dub videos, clone a voice or help people who cannot speak easily.

Good forAccessibility, narration, training material and language dubbing.

Watch forConsent, scams, impersonation and where your voice sample is stored.

Music

Songs, loops and audio ideas

Suno, Udio and similar tools can generate songs from prompts. They are fun, but the legal ground is still moving.

Good forSketching ideas, demos, jokes, learning structure and non-commercial play.

Watch forCommercial rights, style copying, lawsuits and platform terms.

Connectors

AI that can touch your accounts

Connectors let AI read or act across Gmail, Drive, calendars, Slack, Notion, GitHub, Microsoft 365 and other services.

Good forFinding files, summarising inboxes, scheduling and work automation.

Watch forPermissions. A connector may see far more than one prompt.

Agents

AI that takes steps for you

Agents can browse, code, click, call tools, write files or chain tasks. They are powerful because they can do real things.

Good forResearch, coding, admin chores and repeatable workflows.

Watch forWrong actions, file access, account access and unclear responsibility.

World builders

AI-made spaces and environments

Google Genie, World Labs and similar projects point toward prompts becoming interactive places, not just flat clips.

Good forGame prototypes, education worlds, simulation, design and story spaces.

Watch forEarly access, rights to generated worlds, safety, physics and cost.

Beginner path

A simple order to try things.

Start where mistakes are cheap. Move toward private data and automation only after you trust the tool and understand the settings.

1. Ask and check

Use a chat tool to explain something public. Ask it for sources. Check at least one source yourself.

2. Draft and edit

Ask for a draft, then rewrite it in your own voice. Do not send AI text blindly.

3. Make and label

Create harmless images, audio or video. Label generated media clearly when someone could mistake it for real evidence.

Examples to compare

Names you will see in the wild.

This is not every tool on Earth. It is the bigger public shelf a beginner is likely to run into, cross-checked against broad model-landscape references such as LifeArchitect and official provider pages.

Tool shelf	Common names to recognise	Plain-English watchout
Super apps and assistants	ChatGPT, Claude, Gemini, Microsoft Copilot, Meta AI, Grok, Apple Intelligence, Perplexity	Great first shelf, but connectors and account data raise the risk.
Frontier and enterprise LLMs	OpenAI, Anthropic, Google DeepMind, xAI, Mistral, Cohere, Amazon Nova, NVIDIA Nemotron, IBM Granite, AI21, Reka	Model quality is not the same as fair terms or safe data handling.
Open-weight and open-source models	Llama, Mistral open models, DeepSeek, Qwen, GLM, Falcon, OLMo, EleutherAI, Hugging Face, Ollama	Open can still have licence limits, safety issues and hardware costs.
Image	Midjourney, DALL-E, Imagen, Adobe Firefly, Stable Diffusion, FLUX, Ideogram, Leonardo	Rights, public galleries, style copying and fake evidence matter.
Video	Sora, Veo, Runway, Luma Dream Machine, Pika, Kling, Hailuo, Synthesia, HeyGen	Credits vanish quickly and realistic video can mislead people.
Voice	ElevenLabs, OpenAI voice, Google voice tools, PlayHT, Resemble, Cartesia	Consent is the big line. Never clone a real voice casually.
Music	Suno, Udio, Stable Audio, MusicFX, Lyria-style systems	Commercial rights and training-data lawsuits are still unsettled.
Connectors and agents	ChatGPT connectors, Claude integrations, Microsoft 365 Copilot, Google Workspace Gemini, Zapier, Make, Notion AI, Glean, Replit, Cursor	Permissions can expose inboxes, drives, calendars, repos and work history.
World builders	Google Genie, World Labs, NVIDIA Cosmos, game-engine AI tools, open world-model experiments	Early category. Ask about generated-world rights, exports, safety and source worlds.